Download Felix F-750 User Manual

DECLARATION OF CONFORMITY
Manufacturer:
CID Bio Science, Inc.
Felix Instruments – Applied Food Science
1554 NE 3rd Ave
Camas, WA 98607
Declares that the CE-marked Product:
Product Models (s):
Model F-750
Complies With:
89/336/EEC Electromagnetic Compatibility
Directive
73/23/EEC Low Voltage Directive
Compliance Standards:
EN 55027
RF Emissions Information
Technology Equipment
EN 50082-1
EMC Immunity Standard
EN 60950
Safety of Information
Technology Equipment
Including Electrical
Business Equipment
February 11, 2015
Leonard Felix
President
Document Overview
Introduction ............................................................................. 1
Features ............................................................................... 4
Specifications....................................................................... 5
Theory of Operation ............................................................ 6
Operating Instructions.............................................................. 9
Unpacking the F-750............................................................ 9
Loading the Battery ............................................................. 9
Basic Operation ................................................................. 11
Navigating Menus .............................................................. 13
Menu Map ........................................................ 14
Select Model ..................................................... 15
Browse Measurements ........................................ 16
Setup Instrument
............................................... 17
Warnings ........................................................................... 23
F-750 Data ......................................................................... 24
Instructions for Creating a Model .......................................... 27
Step 1: Creating a Training Set .......................................... 27
Step 2: Collecting Reference Values .................................. 34
Step 3: Importing a Training Set and Reference Values into
Model Builder .................................................................... 35
Step 4: Creating a Model ................................................... 39
Step 5: Saving a Model to F-750 ........................................ 51
Step 6: Validating the Model ............................................. 51
Additional Instructions: ..................................................... 53
Merging Training Sets in Model Builder ................... 53
Alternate Method for Training Set Creation: Manual
Data Collection and Entry ..................................... 55
Building a Model to Measure Two Traits .................. 61
Using the Small Fruit Adaptor Accessories
............... 63
Cleaning and Maintenance ................................................ 66
Firmware Update Procedure ................................. 67
Supporting Science ................................................................. 68
Fish Fat and Moisture Content........................................... 68
Wine Grape Anthocyanin Content ..................................... 69
Mango Model Optimal Wavelengths ................................. 70
Kiwifruit.............................................................................. 71
Apple Dry Matter ............................................................... 71
Instrument Precision Optimization .................................... 71
Publications ........................................................................ 72
F-750 Vocabulary Definitions .................................................. 75
Technical Support ................................................................... 79
Customer Service ............................................................... 80
Frequently Asked Questions .............................................. 81
Felix Instruments Hardware Warranty ................................... 85
Appendix I: Mango Model Building Standard Operating
Procedure (SOP)...................................................................... 87
F-750 Production Test Check Sheet ........................................ 94
Warranty Registration Card .................................................... 96
This manual is written for firmware version 1.1
F-750 Instruction Manual 7/30/2015
Introduction
The F-750 Produce Quality Meter aids agricultural
suppliers in developing a fruit maturity and sweetness
management process for the benefit of growers,
processors, and consumers. Judging fruit maturity by
shape and color alone results in varying success. The F-750
delivers the stability, repeatability, and precision that is
necessary
for
successful
chemometric-based
measurements without destroying the product in a
portable easy-to-use device. It comes equipped with a GPS
for spatial mapping, enabling users to generate orchardwide maps of fruit parameters. With this information,
users can plan harvest sequences and improve crop
management techniques. The F-750 uses Herschelinfrared or NIR-a spectroscopy to estimate a variety of
important produce quality indicators such as dry matter,
Brix (total soluble solids), or acid content.
“For quantitative analysis of complex samples,
fast and inexpensive spectroscopic methods are
preferable to the slow, expensive, and destructive
‘wet chemical’ approach. Historically, one
disadvantage of spectroscopic methods has been
the difficulty of determining frequency regions
where the constituent of interest selectively
interacts with light. Now with the power of
multivariate techniques such as Partial Least
Squares, we can use the spectra (X) to predict the
concentration of the constituent of interest (Y),
with accuracy approaching that of wet chemistry.”
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
1
F-750 Instruction Manual 7/30/2015
Adapted from Sjostrom and Wold (1983). A multivariate
calibration problem in analytical chemistry solved by partial
least-square models in latent variables. Analytica Chimica
Acta, (150) 61-70.
Dry matter (DM) is the ratio of the water content to the
dry weight of the fruit. Dry matter is an indicator of both
taste and texture in certain fruits. High dry matter fruits
have sweeter tastes, stronger flavors, higher acidity,
greater quantities of sugar and a higher vitamin C content.
Fruit continue to accumulate dry matter until harvested.
Once harvested, the DM of the fruit and the upper limit for
soluble sugars the fruit develops is fixed. The soluble sugar
content (SSC) of a fruit and the degrees of brix (Obrix) are
interchangeable units of measure for sugar concentration.
In order to generate an NIR (Near Infrared) prediction
model, a calibration or training set must be created by
collecting the spectra of samples of the fruit in interest
using the F-750. The parameter(s) (i.e. dry matter, Brix,
etc) are then collected from the same samples, often a wet
chemical approach, and loaded onto provided software
along with the training set to build the NIR prediction
model. By scanning fruits with the F-750 loaded with this
model, the user can use the information from the device
to determine readiness for harvest, sale, or consumption.
2
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
F-750 Produce Quality Meter
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
3
F-750 Instruction Manual 7/30/2015
Features
 Portable, lightweight and easy to operate.
-One button operation for recording new measurements
 Repeatable, precise, and non-destructive
measurements.
-Spectrometer standard deviation > 0.017%
 True sunlight readable transflective display.
-Contrast of this display actually increases under brighter
sunlight
 Removable, re-chargeable standard sized
batteries.
-Two sets of batteries included
- Stand-alone battery charger included
-Additional button-top 19670 (or protected 18650) batteries
can be purchased from a preferred battery vendor
 48 channel GPS
-Allows spatial mapping of collected data
4
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Specifications
Spectrometer
ZEISS MMS1 VIS-NIR
Wavelength Range
285 – 1200 nm (+/- 10 nm)
Spectral Sampling
3 nm
Spectral Resolution
8-13 nm
Accuracy
Battery Life
Expected: +/- 0.5oBrix RMSEP
(root mean square error of
prediction)
1600 scans (Variable)
PC Interface
USB and SD card
Data Recorded with
Each Measurement
Temperature Range
Raw Data
Reflectance
Absorbance
Second & First Derivatives
Second Derivative Absorbance
0-50OC
Weight
1 kg
Dimensions
7.1 x 4.75 x 1.75 inches ( 18 x 12 x
4. 5 cm )
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
5
F-750 Instruction Manual 7/30/2015
Theory of Operation
In order to generate an NIR model, create a training set
using the F-750 to record 2nd derivative absorbance data.
In order to make a strong model, training set samples must
consist of the entire range of fruit development for the
internal fruit quality parameter to be analyzed. Other
sources of variability should also be considered when
selecting training set samples and should include samples
with any parameters that could affect spectral response.
Next, using these same specimens, a property of interest
(reference value) is collected from the scan location on
the specimen. Typical specimens include produce such as
apples, mangos, or grapes, and a variety of parameters can
be measured such as brix or dry matter.
The Training Set and Reference Values gathered are
loaded into the supplied Model Builder software which
identifies correlations between the Reference Values and
2nd derivative absorbance spectra from the Training Set.
This is accomplished using non-linear iterative partial least
squares (NIPALS) regression. The result is a prediction
model which can be loaded onto the F-750 device and
used to measure the parameters used to build the model.
With a model loaded on the F-750 device, the user places
a specimen on top of the scanner and presses the scan
button to record a new measurement.
6
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Figure 1: Interactance optical design; light rays illuminate an
anulus on the sample. The light then interacts with the sample
by internally scattering through the tissue. Light that undergoes
remission normal to the collimating lens is collected and
focused onto the fiber.
While recording a new measurement, the device…
a. Normalizes the spectrometer output using a
reference shutter.
b. Records darks scans to account for dark current
and ambient light.
c. Calculates diffuse reflectance by subtracting the
light reflected from the reference shutter from
the light reemitted by the subject.
d. Calculates reciprocal absorbance:
Log(1/Reflectance).
e. Calculates first and second derivative spectra by
applying Savitzky-Golay coefficients.
The second derivative spectra is then processed by the
prediction model, each wavelength is multiplied by the
regression coefficient, and all of the wavelengths are
added up. A final intercept coefficient is added, resulting
in the predicted value which the F-750 displays on screen,
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
7
F-750 Instruction Manual 7/30/2015
and saves to the instrument’s SD card along with other
information such as GPS location, battery status, and
temperature.
8
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Operating Instructions
Unpacking the F-750
The F-750 will arrive in a carrying case with removable,
rechargeable batteries, two battery chargers (one for the
wall and a car charger), a white Teflon calibration disc, and
a clip to attach the F-750 to a belt. In addition, an SD card
is included along with a hand strap. There is a tripod mount
located on the bottom of the F-750 Produce Quality Meter
case. Also included is a USB cable which is used to update
the firmware on the device.
Loading the Battery
The F-750 uses 18650 Li-ion 3.7V 3100mAh rechargeable
batteries. The batteries must be removed from the F-750
to be charged. To remove the batteries, twist the battery
compartment cap, located on the bottom of the device
under the rubber bumper. The cap can be twisted with
fingers or a screwdriver to tighten or loosen. Use caution
when removing batteries, as the cap is spring loaded.
Both batteries should be inserted into the unit positive (+)
side first (towards lens side).
Warning: Do not drop batteries, this may cause
them to crack and rupture.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
9
F-750 Instruction Manual 7/30/2015
Top View:
Front View:
Scan Button
Power
Button
Directional Keys
10
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Basic Operation
The rubber bumpers on the top and bottom should always
be installed when using the instrument as they prevent the
device from water damage. If the lens becomes dirty, it can
be cleaned with a soft cloth. The lens is made of extremely
durable Gorilla Glass and should not become scratched
with normal use. The strap should be tightened so that
the device has a snug and secure fit around the hand.
To take a measurement, power on the F-750 Produce
Quality Meter. The current model will be displayed. To
change the model refer to the Select Model section of the
manual. Place the sample on the lens so that it makes
contact with the perimeter of the bracket around the lens.
Press the scan button to record a new measurement.
During the measurement, the instrument will switch to the
scan display, then collect and process the data. As the
measurement is being collected, large blocks will appear
at the bottom of the display screen. Keep the sample in
place as the blocks progress until the measurement is
complete.
The measurement is being collected as the 1st block is
displayed. The 2nd block is for processing data, and the 3rd
block is for saving the measurement. A final screen will
indicate that the data has been saved to the SD card before
displaying the record to be reviewed.
The data is stored to a removable SD card. To remove,
press SD card into the unit and it will eject. To replace,
insert metal side first with the label facing the back of the
unit.
1554 NE 3rd Ave, Camas, WA 98607, USA
11
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
If the instrument locks up or requires a hard shut-off for
any other reason, hold down the power button for 10
seconds. This will force the F-750 power to turn off. Then,
the instrument can be restarted.
When the F-750 device is powered on, it will make a single
beeping sound if it is working smoothly. If not, the device
will issue more than one beep. The number of beeps can
be used to diagnose several errors that the instrument is
able to detect during start-up. The following table outlines
the number of beeps and the associated errors. Please
contact [email protected] if you are unable
to bypass an error.
Number of Beeps
2
3
5
6
7
8
9
10
13
15
16
20
30
12
Error
Device in bootloader assist mode
Battery low
Settings corrupt
Model missing
Spectrometer error
Lamp error
Shutter error
Keystroke error
Graphics error
SD card error
Unhandled error
Unknown board layout/version
RAM (memory) error
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Navigating Menus
Use the power button to turn on the F-750. After the
instrument powers on, the display will show the current
model loaded on the unit and the battery life at the top of
the display. Instructions in the middle of the display
indicate to press the scan button to record a new
measurement or press the right arrow to access the main
menu.
Once at the Main Menu, the left arrow can be used to go
back or exit to the previous screen. The Main Menu
consists of three options: Select Model, Browse Records,
or Setup Instrument.
Menu Navigation Controls
Key
Right Arrow
Left Arrow
Up or Down Arrow
Hold Directional Arrow
Power Button
Scan Button
Function
Enter/engage
Exit/go back/erase
Scroll to select option
Fast scroll to select
number/letter and to
scroll through
measurements
Power on/off
Record a Measurement
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
13
F-750 Instruction Manual 7/30/2015
Menu Map
14
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Select Model
The device will come pre-loaded with demo models. Use
the up/down arrows to highlight the desired selection. To
make modifications to the model, press the right arrow
when the model is highlighted to get to the model submenu.
Load Model: When the right arrow is pressed while this is
highlighted, the current model will be loaded, and the
screen will return to the beginning display.
-Collect Measurement: If the sub-menu of a model that
is already loaded is entered, the first option will be Collect
Measurement, rather than Load Model. Pressing on the
right key with this highlighted will collect a measurement,
similarly if the Scan button is pressed.
Measurements Per Specimen: This option allows the user
to set how many consecutive scans the device takes
automatically after the scan button is pressed.
Delay Between Measurements: If the measurements per
specimen is more than one, the Delay Between
Measurements option allows the user to set a pause of
however many seconds desired in between the
consecutive scans. The pause allows time for the device to
cool off in between scans or may be used by the user to
interrupt the consecutive scans by pressing any of the
arrows and then pressing the left arrow to return to the
Main Menu.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
15
F-750 Instruction Manual 7/30/2015
Scans to Average: This option allows the user to select the
number of scans (measurements) that are averaged
together before a value is shown on the display.
Advanced Options: Several options can be accessed using
the model sub-menu, such as the Integration Time,
Intercept Coefficients, Lamp Off Shutter Open, and Lamp
Off Shutter Closed. The Auto Integration Time option
allows the user to switch between auto integration time
and manual integration time. When it is disabled, the
integration time values can be set manually. Only when
the Auto Integration Time is disabled are the options to
adjust the integration time turning the lamp on and off will
appear.
Using the Intercept Coefficient 1 and Intercept Coefficient
2 options, a new value to use for the intercept coefficient
can be entered. The intercept coefficient is from the
regression vector.
Press the left arrow to get back to the model sub-menu. If
the left arrow is erasing characters on the display, keep
pressing the left arrow to erase all the characters and then
it will exit to the previous menu.
Browse Measurements
Once a measurement is taken, it will be saved onto the SD
card and may be accessed by pressing the right arrow with
Browse Measurements highlighted. Use the up/down
arrows to highlight the desired measurement. Once it is
highlighted, press the right arrow to access the
measurement data. Each measurement has different
parameters saved in it, depending on the model used
16
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
when taking the measurement. For example, a mango
model may include values for dry matter, color, and
chlorophyll.
Press the right arrow while on the measurement data
screen to enter the file menu.
Change Lot #: This option allows the lot number of the
measurement to be changed.
Change Sample #: This option allows the sample number
of the measurement to be changed.
Delete Measurement: This option is used to remove a
record permanently.
Rename Measurement: This option is used to change the
filename of the measurement.
Measurement Details: This menu provides information
like the model used to collect the measurement,
timestamp, file size, collection details, and location of the
scan.
Setup Instrument
The Setup Instrument menu houses the features and
settings that are adjustable by the user.
Lot & Sample #: The lot and sample numbers are used to
help organize data so that the groups from which samples
are taken from may be recorded. For example, if multiple
boxes of apples had to be scanned, the lot number would
be assigned to each box and the sample number would be
assigned to each apple in a box. The lot and sample
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
17
F-750 Instruction Manual 7/30/2015
numbers are included in the filename of a measurement
and increment with each scan.
-Current Sample #: This option is used to change the
sample number for the next scan.
-Current Lot #: This option is used to change the sample
lot number for the next scan.
-Increment/Decrement Logic: This option is used to
enable or disable the decrement if measurement deleted
option, which if enabled will automatically decrease the
sample number by 1 if the latest measurement is deleted.
It also includes the Increment Lot # Every x Samples option
which is used to set the number of scans before the lot
number is increased by 1 and the sample number is reset
to 1. The lot and sample number system may be enabled
with the Append As Measurement suffix option, or
disabled, upon which the first scan taken will be assigned
the number 1 and increment for each scan after.
Measurement Prefix: A prefix is the label placed on each
measurement along with the lot and sample number that
may help with organizing data. Within the Measurement
Prefix option, a new prefix can be entered or a recently
used prefix can be selected. The New Prefix option allows
the user to choose a new prefix. The prefix has a maximum
of 16 characters. The Select A Recently Used Prefix option
will generate a list of recently used prefixes. Use the
up/down arrows to highlight the desired prefix to be used,
and the right arrow to select the desired prefix.
18
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Archive Measurements: The Archive Measurements
menu will store all data currently in the populating the list
of measurements into an archive folder on the SD card.
The data in the archive folder may now only be accessed
through DataViewer. The Create Folder option will make a
new archive folder with the timestamp as a default name.
The name may be changed and upon confirming the name
by pressing the right arrow at the end of the name, all data
in the browse measurement menu will be transferred to
the folder. The Select Folder option allows all data in the
browse measurement menu to be transferred into a
previously created archive folder by moving the cursor
with the up and down arrow and pressing the right arrow
when the cursor is highlighting the desired archive folder.
Date & Time: The Date & Time option allows the
timestamp on the F-750 to be adjusted.
Training Sets: Creating a training set is the first step in
creating a model. This menu is used when collecting
spectra to use in the training set.
-Create New Training Set: First, enter the name to use
for this new training set. Next, enter the number of
specimens that will be used to build this model. The
minimum number of specimens required is 10 and the
maximum number allowed is 5,000.
Once a new training set is created the options are Browse
Specimen Data and Delete Training Set. If the Browse
Specimen Data option is chosen the user is prompted to
choose from a temperature range (minimum, mid, and
maximum). At each temperature, there is a numbered list
1554 NE 3rd Ave, Camas, WA 98607, USA
19
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
of the specimens. Highlight a specimen by using the
up/down arrows. When Specimen 1 is highlighted, press
the right arrow to collect data for the training set. The
instrument will indicate “Processing” and then “Creating
Training Set”. Once it has completed, do not press the scan
button to gather training set spectra. The scan button only
initiates a measurement. The space next to Specimen #1
will change from (empty) to (timestamp). Repeat with
each of the 20 specimens at each of the 3 temperatures.
Specimens can be re-done if a mistake is made or the
wrong specimen/temperature is selected.
-Modify Existing Training Set: This menu produces a list
of training sets on instrument. Press the right arrow on a
desired training set edit, or delete a training set.
GPS Receiver: The GPS receiver menu shows the state of
the GPS sensor as enabled or disabled. If the GPS sensor is
enabled, View Status will appear with the following
parameters: latitude, longitude, satellites in view, whether
there is a satellite lock or not, and Horizontal Dilution of
Precision (HDoP). When taking a GPS reading, be sure to
keep the unit upright and the top of the instrument free
from obstruction.
Keypad Sounds: This option allows the noises the
instrument makes when the arrow are pressed to be
changed. There are four options: Disabled, in which the
device will make no sound when a key is pressed, and then
Quiet, Default, and Loud.
Diagnostics: This menu is intended for debugging and
support. Within the Diagnostics>Manual Test menu, the
20
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
following options are available: Scans to Average,
Integration Time, Lamp State, Shutter State, Number of
Repetitions, Delay Between Repetitions, Measure
Spectrometer Noise, and Collect Spectra.
-Scans to Average: This option allows the user to select
the number of scans (measurements) that are averaged
together before a value is shown on the display. The
integration time for the calibration-collect spectra can also
be set.
-Lamp state: This option is used to turn the lamp on or
lamp off. Press the down or up arrows to turn the lamp off
or on. The shutter state works similar to the lamp state.
The options are shutter closed or shutter open (press
down/up to open/close shutter).
-Collect spectra: This option will use whatever values are
set in the diagnostics menu to collect spectra. After
rendering, the results will be saved to the SD card in .raw
format. This data can be analyzed by inserting the SD card
into a computer and extracting the data.
Advanced Setup: If the advanced setup menu is selected,
a warning will appear that calibration should only be
performed by authorized users. Press right to continue.
Press left to abort. Instruments should only be calibrated
under direction of a Felix Instruments technician. The
options in this menu allow the instrument to be specifically
calibrated for optimum performance. Many of these
values are instrument specific such as pixel coefficients
and reference voltage. The pixel coefficients are the
spectrometer calibration values given by manufacturer
1554 NE 3rd Ave, Camas, WA 98607, USA
21
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
(C0, C1, C2, C3). The pixel coefficients influence what
wavelengths of light the spectrometer is responding too.
-Spectrometer Zero and Span: This option is used in
conjunction with the white Teflon calibration standard.
After the instrument has been placed in a cold
environment for at least 30 minutes, place the white
Teflon calibration standard over the lens and press the
right arrow to take a scan. The display will flash “Saving…”
and exit to the menu when the zero offset is complete.
In addition to spectrometer settings, lamp, shutter,
temperature, and auto integration time settings can also
be fine-tuned under their corresponding menus.
Factory reset: This option will erase the temporary cache
files for the display screens and return to the default
parameters from the manufacturer.
22
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Warnings
 Do not drop! Dropping the instrument while it is
turned on could cause the lamp bulb to break.
 Instrument will become warm with use, due to the
lamp turning on.
 Do not drop F-750 batteries on hard surface, such
as cement floor.
 Do not leave the lamp on for more than 30
seconds. If the lamp remains on, power off the
instrument or remove the batteries.
Dropping the batteries on a hard surface may cause them
to crack the seal and rupture. The electrolyte in the
battery will slowly leak out; the electrolyte is toxic and
may cause burns.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
23
F-750 Instruction Manual 7/30/2015
F-750 Data
Data collected by the F-750 can be reviewed on the
instrument by using the Browse Measurements option.
The data is stored on the removable SD card. To remove,
press the SD card into the unit and it will eject. To replace,
insert metal side first with the label facing the back of the
unit. It is best to eject and insert the SD card when the
instrument is off to prevent file system errors on the SD
card.
In order to view the files on a computer, the F-750 Data
Viewer software should be installed. This software can be
found on the Felix Instruments webpage located at:
http://felixinstruments.com/support/f-750support/software
Individual or multiple training sets of measurement scans
can be opened in the DataViewer by going to File>Open.
24
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Once a file or files have been opened in DataViewer, data
can be exported to be viewed. The File>Export menu gives
four different options to export: Measurements,
Measurements & Details, Raw Spectra, and Interpolated
Spectra.
To export the predicted values, select Measurements and
choose to open the newly created .csv file. During the
export, the software will prompt for a save location and
again after the export is successful to open the newly
created file.
The Raw Spectra and Interpolated Spectra export features
are mainly for use by technicians. The interpolated spectra
menu will allow the spectra to be accessed.
The Archive feature will extract all possible data from the
selected files and save as CSV.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
25
F-750 Instruction Manual 7/30/2015
After exporting, the .csv file can be opened as a
spreadsheet. This .csv can be used in other software
programs. Data such as model results, which contains GPS
coordinates, dry matter, TSS, battery life, and raw readings
can be viewed in the .csv file. Other relevant data includes
the interpolated 2nd derivative absorbance data, which
contains what the model is plotting and the value at each
wavelength.
26
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Instructions for Creating a Model
Step 1: Creating a Training Set
Step 2: Collecting Reference Values
Step 3: Importing a Training Set and Reference Values
into Model Builder
Step 4: Creating a Model
Step 5: Saving a Model to F-750
Step 6: Validating the Model
Additional Instructions:




Merging Training Sets in Model Builder
Alternate Method for Training Set Creation:
Manual Data Collection and Entry
Building a Model to Measure Two Traits
Using the Small Fruit Adaptor
Step 1: Creating a Training Set
For training set creation, use pieces of fruit across the full
range of fruit maturity. In general, the more specimens
used to make the model, the more accurate it will be at
predicting values. Do not include any fruit that has
suffered from heat stress, sun burn, is misshapen or
otherwise physically damaged.
Influence of Temperature on Spectra: The purpose of
using multiple temperatures in training set creation is to
compensate for the changing molecular response to light
in relation to temperature. By scanning the same fruit at
2-3 temperatures, the program is able to ignore any
spectral shifts or changes that do not relate to changes in
the desired trait such as fructose or starch. By
1554 NE 3rd Ave, Camas, WA 98607, USA
27
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
including variation of temperature for the same fruit in the
model, the math can autocorrect for temperature. For
some applications the need for temperature correction is
small, such as in controlled storage where temperature is
constant, but if the F-750 will be used on fruit which does
not always remain a constant temperature, a training set
will need to be created to accommodate for this potential
temperature range.
If, at any point during the training set creation process, you
navigate away from the training set you are working on or
turn the instrument off, you can navigate back to the
training set by going to Setup menu > Training Sets >
Modify Existing Training Set, then selecting the training set
you wish to continue working on.
The following instructions are for creating a model at three
temperatures:
1. Select pieces of fruit with a wide range of
maturity to be used in training set creation.
28
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
2. Label each piece of fruit with a specimen
number.
3. Mark the area to be scanned on each fruit.
Two sides of the same fruit are often scanned
depending on the variability of the fruit (sun
side and shade side). If this is the case, label
both sides with separate specimen numbers.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
29
F-750 Instruction Manual 7/30/2015
4. Determine the three temperatures to be used
for the training set (Minimum, Mid, and
Maximum). It is best to choose a range of
temperatures encompassing the range which
the measurements will be taken, for example,
10ºC, 20ºC and 30ºC.
5. Place the fruit in a temperature controlled
area to bring the fruit to the first temperature;
Minimum.
6. After the fruit has been given enough time to
come to the pre-determined temperature
(generally about one hour), it is now time to
begin taking measurements for the training
set.
7. Turn on the F-750 and navigate to the Setup
Menu > Training Sets > Create New Training
Set.
8. Enter a name to use for the new training set.
Press the right arrow to enter.
9. Enter the number of specimens that will be
used to build this model (minimum of 10
specimens, recommended ~200 specimens).
Press the right arrow again to create the
training set file. Be aware this step can take a
few minutes, and very large training sets
(5,000 specimen) can take up to an hour to
process.
10. After the training set has been created, press
right to enter the Browse Specimen Data
menu. Navigate to the first temperature:
Minimum. Now we will begin scanning each
30
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
pre-labeled specimen. Be careful not to
remove too many specimens at a time from
the temperature controlled environment, as
you want the set temperature to remain
constant across specimens.
11. Place the first specimen on the lens, and with
Specimen 1 highlighted on the list, press the
right arrow to collect the training spectra.
The instrument will indicate “Processing”. Do
not press the square/scan button to gather
training set spectra.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
31
F-750 Instruction Manual 7/30/2015
12. After each specimen has been scanned, you
will see a date and time included in the
specimen file name. This is how you know the
data has been recorded. Specimens can be rerecorded if a mistake is made or the wrong
specimen/temperature is selected.
13. Repeat for all specimens at minimum
temperature.
14. After all specimens have been recorded at
minimum temperature, place the specimens
back in the temperature controlled area to
bring the fruit to the next selected
temperature; Mid Temperature.
15. After the specimens have reached the desired
mid temperature, the next set of temperature
data is ready to be collected. Always scan the
same area of the fruit.
16. If you had turned the F-750 off while waiting
for the specimen to reach their next
temperature, navigate to the Setup menu >
Training Sets > Modify Existing Training Set,
and select the training set you created
previously.
17. Within the training set file go to Browse
Specimen Data and navigate to the Mid
Temperature folder. Place the first specimen
on the lens, and with Specimen 1 highlighted
on the list, Press the right arrow to collect the
training spectra.
32
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
18. Repeat for all specimens at mid temperature.
(Note: You will be using the same specimens
for all three temperature sets so the specimen
must be scanned in the same order each
time.)
19. After all specimens have been recorded at mid
temperature, place the specimens back in the
temperature controlled area to bring the fruit
to the last selected temperature; Maximum.
20. See step 16 if you have turned the F-750 off
while waiting for the
21. Wait for the specimen to reach their next
temperature, then continue to next step.
22. Within the training set file go to Browse
Specimen Data and navigate to the Maximum
Temperature folder. Place the first specimen
on the lens, and with Specimen 1 highlighted
on the list, Press right arrow to collect the
training spectra.
23. Repeat for all specimens at maximum
temperature.
24. You are now finished creating your training
set. The same specimen used for training set
creation will now be used to collect reference
values in Step 2 “Collecting Reference
Values.”
25. It is recommended to import data into Model
Builder software to verify scans are saved
before destructive sampling of fruit.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
33
F-750 Instruction Manual 7/30/2015
Step 2: Collecting Reference Values
To collect Reference Values, the same fruit specimens
used in training set creation will be measured. The
Reference Values are those values determined by your
reference method (i.e. a destructive measurement of TSS
or DM). For example, if you intend to measure brix,
Reference Values could be measured with a
refractometer, or total acids could be measured through
acid titration. Careful data management is paramount for
this step, and extra care should be taken to keep
specimens well organized and numbered. For the
purposes of this example, we will describe the method for
measuring dry matter.
1. First, slice off the fruit cheek section which was
marked and scanned with the F-750.
2. Remove the fruit skin from the sample with a
sharp knife or peeler.
3. Cut the sample into a small piece, skinless piece.
An average piece would be about the size of a
2x2x2 cm cube or smaller, depending on the size
of the sample.
4. Weigh the sample as soon as possible and record
the weight to the nearest 0.001 grams.
5. Dry the samples in a dehydrator for 48 hours set
at 63-65ºC.
6. After 48 hours, re-weigh the samples and record
the values. Dehydrator times and temperatures
may vary for different types of fruit.
7. Using these weights, dry matter (%) can be
calculated for each specimen by dividing the dried
34
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
mass by the original mass. These dry matter
values will be our Reference Values
corresponding to the measurements taken in the
creation of our training set.
Step 3: Importing a Training Set and
Reference Values into Model Builder
After both training set data and Reference Values have
been collected, we now have the necessary data to build a
model. It is important to save your Model Builder file so
that you can return to it at a later time and add additional
data if desired.
1. Open the Model Builder software program. If you
have not already downloaded this program onto
your
computer,
please
download
at:
www.felixinstruments.com/F750/ModelBuilder
2. Remove the SD card from the F-750 and insert it
into your computer.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
35
F-750 Instruction Manual 7/30/2015
3. First, we will import our training set. In Model
Builder, navigate to the Import button on the right
side of the software window.
4. You will be prompted for the location of the
training set file. To select the desired training set,
navigate to the SD card folder on your computer,
click on training sets, and select the training set
you created. Click Open.
5. You will be prompted with the message “Would
you like to overwrite the existing training set?”
Select Yes If interested in adding to an existing
Model Builder project, please go to the Adding
Data to an Existing Training Set instructions found
on page 53.
6. When you created your training set, if any
specimen files were left empty, they will
automatically be excluded.
7. Model Builder will automatically populate the
training set and the spectra should now be
36
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
displayed on screen. Specimen number, Spectra
Name (Min, Mid, Max temp), and the pixel values
can be found under the Training Set tab at the
bottom of the screen.
8. Use the mouse wheel to zoom in/out on the graph.
Click and hold the graph to pan.
9. To select a single specimen in the spectra, select
the row in the training set list below, or hold the
control key (Ctrl) and click to choose the line in the
graph above. To select all specimen rows, click the
empty cell to the left of the Exclude column
header. This will select all specimens imported in
the training set. To exclude a specimen, click the
Exclude box to the far left of each specimen row.
Multiple rows can be selected by holding down
the shift key for a contiguous set or the ctrl key
for individual rows.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
37
F-750 Instruction Manual 7/30/2015
10. Next, we will import our Reference Values. If they
are in CSV format, click the Import button. If they
are not in CSV format, they can be manually
entered or copied and pasted from Excel or similar
spreadsheet programs as a column with no
header. Specimen numbers are auto-populated
starting at 1.
11. Be sure that all specimens included in the training
set also have corresponding and matching
Reference Values. To change the column headers,
38
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
double-click to rename them (e.g. Dry Matter or
TSS). Model Viewer can detect if there are missing
or multiple specimens.
Step 4: Creating a Model
Now that we have both our training set data and our
Reference Values imported into Model Builder, we are
ready to build or create the model. Model Builder
illustrates spectra quality (noise). It is recommended that
the user manually selects regions to include by choosing a
spectra range, such as 729-975 nm.
1. Model Spectra Window Selection
a. Generally, it is best to manually select a
spectra range based on the known range
of a particular fruit or constituent being
measured. For example, choosing a range
of 729-975 nm for sugar or starch. For
additional ranges, see the “Supporting
Science” section at the end of this manual.
To manually select a region, enter the
desired range under the Model Spectra
tab located at the bottom of the window.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
39
F-750 Instruction Manual 7/30/2015
b. Seeing a neatly layered rainbow effect
with the colors (green-orange-red) in the
selected range is optimal and indicates
agreement of spectra and known results.
2. When all data are entered and modifications
made, build the model by clicking the Build button
in the toolbar near the top of the window. The
application will indicate that the model has been
successfully built and a new tab will appear at the
bottom of the screen: Model Performance
Analysis.
Alternate method:
The Model Builder software has a feature to suggest the
useful region of spectra. This feature is based on univariate
correlation per pixel. It is important to note that selecting
only pixels with a high univariate correlation does not
always translate into good multivariate regression and
therefore a robust model. It is important to have enough
points to define a curve and peaks are 20-50 nm wide.
40
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
1. To begin the model building process navigate
to the Find Correlation button on the right
side of the screen.
2. Model Builder will now highlight useful
spectra regions. The red areas represent a
detected correlation and the blue areas are
the wavelengths currently included.
a. Wavelengths to exclude can be
entered into the box on the right side
of the screen. This pane also allows
you to select spectra range by
associated Minimum R2 value. The R2
value represents the strength of the
correlation. High R2 values typically
indicate a strong correlation. Use
caution, setting an R2 value too high
could also eliminate potentially useful
data.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
41
F-750 Instruction Manual 7/30/2015
3. When all data are entered and modifications
made, build the model by clicking the Build
button in the toolbar near the top of the
window. The application will indicate that the
model has been successfully built and a new
tab will appear at the bottom of your screen:
Model Performance Analysis.
Model Performance Analysis
The Model Performance tab allows for the interpretation
and optimization of the underlying mathematical
regression used by the F-750. The various graphs in the
drop down menu allow for analysis of different aspects of
the multivariate data, including principle components
(explained variance), root mean square error (RMSE),
model linearity, prediction error, regression coefficients,
error/deviation ratio, and scores plot. Analysis of
regression coefficients and prediction error will indicate
the fit of the model.
The two major concerns when inspecting a model’s results
for how the regression fits are: number of principle
components and outliers. The optimum number of
principle components to include will be selected by the
program, but alternate numbers of principle components
can also be selected to optimize data fitting. By selecting
too few principle components there will not be enough
representation, too many components and the correlation
could be a coincidence because noise is included.
42
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
The prediction error graph is often the most useful; the
closer the Reference Value is to the prediction, the lower
the prediction error and the more accurate the model.
Using the Model Performance Analysis Tab:
The following outlines each screen or graph seen on the
model performance analysis tab as an interactive
walkthrough using the Grape Demo model that is included
with the F-750 device.
After the model is successfully built, the Model Builder
software will show the Model Performance Analysis tab.
Click on the tab and the first screen shown is the explained
variance.
The explained variance graph is useful for understanding
how well the variation in the known value is spanned by
the regression model. As the explained variance increases,
the model is better incorporating the entire data set. The
information is displayed in a line graph to show the
contribution from each principle component (PC).
Overfitting can occur if principle components are included
in the model that do not incorporate more of the dataset.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
43
F-750 Instruction Manual 7/30/2015
Navigate to the next graph in the drop down list, root
mean square error (RMSE). The blue line indicates the
root mean square error calibration (RMSEC), which
represents the average error of the complete regression.
The red line indicates the root mean square error of cross
validation (RMSECV). This is the average error of a
simulated independent data, either with full or leave-oneout cross validation.
RMSECV is the average error and is considered to be the
error within the first standard deviation or 68% of all
readings. The Model Builder software determines the
optimum number of principle components by looking for a
minimum decrease of at least 5%. In the example below, 8
principle components have an RMSECV of 0.61, a 5%
decrease in error would be 0.58. Because 9 principle
components have an error of 0.59, Model Builder has
determined that 9 and above principle components are an
over fit.
44
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
The following graph in the model performance analysis tab
is model linearity. Model linearity represents the
goodness of fit of the regression. A model with an R2 less
than 0.7 will likely not meet performance expectations.
The Prediction Error graph displays the reference value vs.
the predicted value for both the complete regression and
the cross validation. This graph is useful for determining
outliers or specimen that do not fit well within the
regression.
It is helpful to go back to the reference field and change
the reference value for a scan to see how that influences
RMSECV and the prediction error graph.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
45
F-750 Instruction Manual 7/30/2015
After the prediction error is the Regression Coefficients
graph. The regression coefficient values are applied in a
dot product with spectra to determine the predicted
value.
Navigate to the Scores Plot graph, which is used to identify
clusters or groupings in data. In the example below, the
red table grapes form a small cluster below the green
grapes. The figure below has a red and green circle
indicating the red and green grape clusters. Additionally,
spectra such as the single scan in the lower right quadrant
may be considered a potential outlier as it is far from other
points.
46
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
The Wavelength Influence graph shows how much each
individual regression coefficient is contributing to the
predicted value, essentially indicating the important pixels
for prediction.
Optimizing the Grape Demo Model:
Now we will go back and optimize the grape demo model.
To begin, navigate back to the Scores Plot and hover over
the Potential Outlier scan in the lower right quadrant. We
can see that this is specimen number 239 with the spectra
name of r39. This stands for the 39th red grape scanned in
the training set.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
47
F-750 Instruction Manual 7/30/2015
To investigate if this specimen is an outlier, we need to go
back to the training set tab and locate specimen 239. Next,
check the Exclude box next to #239.
Now that Specimen 239 is excluded, we can re-compute
the model by pressing the Build icon. Then, explore the
impact of removing specimen 239 from the training set.
Because the RMSE and Model Linearity graphs show no
change, it can be concluded that specimen 239 is not an
outlier and should be included in the training set. Go back
to the training set, include #239 and build the model
again.
Next, look at the Prediction Error graph. Here, you can see
that specimen 260 is poorly predicted by the model.
Repeating the steps above, Exclude specimen 260 and
explore any changes to model performance in the RMSE
and model linearity graphs.
48
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
After excluding specimen 260 and re-building the model, a
decreased RMSE and slightly increased R2 value is seen.
This indicates that specimen 260 is likely an outlier or at
the very least a poor fit for the model.
It is important to note that outlier removal must be done
delicately and for sufficient reasons. It is inappropriate to
simply remove specimen until you have a “perfect model.”
Outliers can result from internal defects, human error in
reference method, or insufficient sampling of real variance
in fruit population.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
49
F-750 Instruction Manual 7/30/2015
Additional Information from Scores Plot:
The following shows the effect of temperature on the O-H
peak (spectra) of a cherry. The upper curve in pink is room
temperature and the lower blue curve is cold storage (min
temp). Temperature alters the photoactive response of
certain common chemicals, like fructose or starch and this
can be seen in the gathered spectra.
50
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Step 5: Saving a Model to F-750
1. To save the model, go to File > Save. Select a
location to save the file on your computer. Now
that the model is saved, you can return to it at a
later time to add additional values or edit the
model.
2. To save the model to an F-750 SD card, go to File >
Save. Locate the SD card and save your new model
in the Models folder.
3. Insert the SD card back into the F-750. Your new
model is now ready to use!
Step 6: Validating the Model
After a new model has been created and loaded onto the
F-750, it is important to validate that the model is working
correctly. To do this, there are a number of different
methods available. One of the simplest methods for on
device validation is as follows:
1. Start the F-750 instrument.
2. Go to Models > select the model just saved to the
device.
3. Press the right arrow and then select Set As
Current Model.
4. Place a specimen on top of the F-750 and press the
scan button.
5. Scan several specimens and measure to gauge the
variance of the model.
6. You can then use your selected reference method
(used to calculate Reference Values) to verify that
these values are accurate.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
51
F-750 Instruction Manual 7/30/2015
On Device Validation: 7 Days Turnaround
This procedure allows for creating and validating a model
for dry matter, titratable acid, or other reference values
with a lengthy process. This procedure is specific for dry
matter.
 Day 1: Training Set scans, prepare and weigh
samples, load dehydrator
 Day 2: Dry (wait time)
 Day 3: Finish drying, final weigh and calculate dry
matter. Build and optimize model in Model Builder
software.
 Day 4: Load model onto F-750 device, take scans
of validation fruit. Prepare and weigh samples,
load dehydrator
 Day 5-6: Dry (wait time)
 Day 7: Finish drying, final weigh and calculate dry
matter. Check predicted value vs. known value on
device.
The procedure for Brix is shorter, with day 1 consisting of
the training set, reference values and creating the model.
Day 2 consists of on device prediction and reference
values.
Notes:

52
The model’s margin of error is based on the
training set, quality of reference method, and if
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015


the region selected in the training set has a
correlation.
Specimens above or below the range of
Reference Values used to create the model will
not predict accurately. This is why it is so
important to select a wide range of fruit at various
stages of maturity when you are creating the initial
training set.
Model performance reflects how well the training
set represents prediction subjects.
Additional Instructions:
Merging Training Sets in Model Builder
Data can be added to an existing Model Builder file at any
time to increase the accuracy of the model or increase the
range of values. Once you have a pre-existing model, to
add to the model you must collect more data with a new
training set and Reference Values. To do this, follow the
same steps outlined above in the Creating a Training Set
(Step 1) and Collecting Reference Values (Step 2)
instructions. We will assume for the following instructions
that these data have already been collected.
1. Open the Model Builder file you wish to edit. You
should see the previous training set and Reference
Values are already loaded in the program.
2. Import the new training set you wish to include. In
Model Builder, navigate to the Import button on
the right side of the display.
3. You will be prompted for the location of the
training set file. To select the desired training set,
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
53
F-750 Instruction Manual 7/30/2015
navigate to the SD card folder on your computer,
click on training sets, and select the training set
you created. Click Open.
4. You will be prompted with the message “Would
you like to overwrite the existing training set?”
Select No.
5. You will be prompted with another message
“Would you like to increment specimen numbers
automatically?” The answer here depends
whether the training set is from the same
population. See the following examples A and B:
a. If the training set #2 is of the same fruit,
click No. Example training set #1 has only
the minimum temperature scans of
population 1 (n=20) and training set #2
has only the middle temperature scans of
population 1. Select No to align the
specimen numbers: min temp fruit_1
should have the same specimen number
as mid temp fruit_1 because they belong
to the same population. This results in 40
total scans with two scans for each
specimen number.
b. If the training set #2 is of different fruit,
click Yes. Example training set #1 has
only the minimum temperature scans of
population 1 (n=20) and training set #2
has only the middle temperature scans of
population 2 (n=20). Select Yes because
the specimen from population #2 should
have unique specimen numbers. The
54
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
training set #1 will have specimen 1-20
ad training set #2 will have specimen
number 21-40.
6. Any empty records will automatically be excluded.
The new training set data should now appear in
the rows beneath the original training set, and all
spectra should be displayed. The new training set
specimen number will start where the first training
set specimen numbers ended. For example, if
training set #1 contains specimens 1-50, when you
import training set #2, the first specimen will start
at 51.
7. Import the new set of Reference Values. You can
also import data files (button scans) and training
set (right arrow scans). If they are not in CSV
format, they can be manually entered or copied
and pasted from Excel.
8. Be sure that the Reference Value specimen
numbers match with the training set #2 specimen
numbers. As with the example from step 6, if your
training set #2 specimen numbers begin at 51,
Reference Value specimen numbers also need to
begin at 51 for training set #2 Reference Values.
Alternate Method for Training Set Creation:
Manual Data Collection and Entry
An alternate method for creating a training set is also
available. This method involves taking measurements with
the scan button (rather than using the training set
feature), then importing the spectra from these
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
55
F-750 Instruction Manual 7/30/2015
measurements into Model Builder using copy and paste.
This alternative method gives more flexibility in importing
data, but extra care must be taken in data organization,
and data entry into the Model Builder software.
First we will collect the data using the manual method:
1. Follow steps from Part 1: Creating a Training Set,
#1-6.
2. Turn on the F-750 and navigate to the Select
Model menu. Select any model. For the purposes
of this method of data collection, the model used
does not make a difference in the spectra
collected.
3. Change the measurement prefix name to keep the
data organized. Go to Setup Instrument >
Measurement Prefix > Enter New Prefix.
4. Now begin collecting data. Place the first specimen
on the lens and press the scan button.
5. Repeat for all specimens at Minimum
Temperature.
6. After all specimens have been recorded at the
minimum temperature, place the specimens back
in the temperature controlled area to bring the
fruit to the next selected temperature; Mid
Temperature.
7. After specimens have reached the desired Mid
Temperature, the next set of temperature data is
ready to be collected.
8. Change the measurement prefix to indicate you
are now taking Mid Temperature data. Go to
56
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
9.
10.
11.
12.
13.
14.
15.
Setup Instrument > Measurement Prefix > Enter
New Prefix.
Place the first specimen on the lens, and press the
scan button to collect the measurement.
Repeat for all specimens at Mid temperature.
(Note: You will be using the same specimens for all
three temperature sets so the specimens must be
scanned in the same order each time.)
After all specimens have been recorded at Mid
temperature, place the specimens back in the
temperature controlled area to bring the fruit to
your last selected temperature; Maximum
Temperature.
Change the measurement prefix to indicate you
are now taking Maximum Temperature data. Go
to Setup Instrument > Measurement Prefix > Enter
New Prefix.
Place the first specimen on the lens, and press the
scan button to collect the measurement.
Repeat for all specimens at Maximum
temperature.
You are now finished creating your training set.
The same specimen used for training set creation
will now be used to collect reference values.
Next, collect your Reference Values
1. Follow steps from part 2: Collecting Reference
Values (See Step 2 above).
Next, extract the F-750 specimen spectra data using Data
Viewer software and import into Model Builder.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
57
F-750 Instruction Manual 7/30/2015
1. Open the Data Viewer software program. This
software can be found the on the Felix
Instruments webpage located at:
http://felixinstruments.com/support/f-750support/software
2. Remove the SD card from the F-750 and insert it
into your computer. Locate the specimen
measurements have taken for the training set.
3. Choose files to open in Data Viewer.
4. Go to the File>Export>Interpolated Spectra. Click
on the menu and select 2nd Derivative. This is the
spectra that we will use for our training set.
5. You will be prompted for a location to save the
data. Select a location, and click save.
6. You will then be prompted with “Process
complete. Would you like to open the created
files?” Select yes.
7. You should now be viewing your data in an Excel
worksheet. The first nine columns include the
measurement prefix, time and date, model used,
shutter open and closed integration time, and
58
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
pixel coefficients. The remaining columns include
the spectra which we will use for training set
creation (285-1200 nm).
8. F-750 scans taken with the scan button can be
imported into Model Builder using the Import
button. Alternatively, scan details can be copied
and pasted into the program. To do this, exclude
the headings in row one, copy all spectra
information. (Be sure you are only copying spectra
data, and that you are copying the entire spectra
285-1200 nm).
9. Open the Model Builder software program. Note
that under the training set tab, the spectra begins
at the same pixel; 285 nm. This is where we will
paste our data.
10. Click to select the 285 nm cell for specimen # 1 in
Model builder. Be sure that it his highlighted blue
as shown in the image above. Now paste the
spectra information copied from Excel.
11. You should see your data automatically populated
for each specimen.
12. Enter the Specimen # and Spectra Name for each
record. You can do this either by manually typing
them into Model Builder, or to expedite this step
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
59
F-750 Instruction Manual 7/30/2015
for larger data sets, specimen numbers can be
copied and pasted.
13. Next, we will import our Reference Values. If they
are in CSV format, navigate to the Import button
on the right side of the Model Builder screen. If the
reference values are not in CSV format, they can
be manually entered or copied and pasted from
Excel.
14. Finally, now that training set and Reference
Value data have been imported, you may
proceed with model building as described in Step
4 “Creating a Model,” Step 5 “Saving a model to F750,” and Step 6 “Validating the Model” sections
of this guide.
60
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Building a Model to Measure Two Traits
Building a model to measure more than one trait (such as
dry matter, brix, or chlorophyll), is often a very useful and
timesaving method. To do this, we will slightly alter our
reference collection method to accommodate the
collection of more than one parameter. To measure two
traits we recommend using one of the following two
methods:
1. Build one training set for one trait and
another separate training set for the second
desired trait. Each training set will use a
different set of specimen.
a. For example, use 100 apples in
training set ‘a’ and collect dry matter
reference values for these specimens.
Use another 100 apples for training
set ‘b’, but this time collect Brix
reference values for these specimens.
Combine these two training sets in
Model Builder as discussed in the
‘Merging Training sets in Model
Builder’ section of this user manual.
2. On fruit which has a large enough area, two
cheeks of the same specimen can be used for
two difference reference methods. To do this,
create a training set by scanning two cheeks of
each specimen.
a. Each specimen should be scanned
twice, once per side. Then, when
collecting reference values, one cheek
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
61
F-750 Instruction Manual 7/30/2015
could be used to measure dry matter
for example and the other cheek Brix.
When building a training set with two scans per piece of
fruit (left cheek and right cheek), care must be taken as to
how the data is arranged in Model Builder software.
When put into model builder with a single training set,
appropriately assign the reference values to the spectra,
as shown in the following figure, where the left cheek is
assigned odd number specimen scans and the right cheek
is even number specimen scans.
62
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Using the Small Fruit Adaptor Accessories
Small fruit adaptor accessories are included with the F750. These accessories are designed to maximize and focus
the light when measuring small fruits such as grapes or
cherries. Two reflector cone sizes are provided to
accommodate a variety of small fruit, 11 mm and 19 mm,
and a small fruit adaptor, which has three spokes and a
central ring, similar to crosshairs or a target.
Fig. Clockwise from Top Left: 11 mm reflector cone, 19 mm
reflector cone, small fruit adaptor.
For very small fruit there is a limited amount of light on the
fruit and the reflector cone can change the angle and
provide more light. For example, a blueberry looks brighter
when placed on the F-750 with proper reflector cone, and
looks very dark without it.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
63
F-750 Instruction Manual 7/30/2015
Selecting a Small Fruit Adaptor Accessory:
It is important to be consistent when taking scans for
training sets and predictions: the accessory should be
used for all samples or for none. If the fruit does not fit in
the reflector cone, and protocol is to use the reflector
cone, pick a different fruit specimen that fits, or re-scan all
fruit samples without the reflector cone.
Ensure that fruit samples are positioned with the same
orientation or presentation when scanning (straight up or
not leaning against the side of the reflector cone). If fruit
touches side of reflector cone, it should touch the side of
the cone for all measurements. Fruit should be centered
using the guides on the small fruit adaptor.
Some cherry varieties may be too large to use the 19 mm
reflector cone, if this is the case, use only the small fruit
adaptor. For cherry varieties, with a diameter typically
larger than 25 mm, it is best not to use a reflector cone or
small fruit adaptor base. Also, extra care should be taken
for these large cherries to take the reference value from
the area of the fruit that is scanned by the F-750.
Keep in mind that consistency in model building is key,
either use the reflector cone for all cherries scanned, or
none.
64
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
To use the reflector cones, one quick changeover must first
be made to the F-750 by installing the small fruit adaptor
(black ring with central circle and 3 spokes). Using a #1
Phillips head screwdriver, swap this adaptor with the plain
one currently on the F-750. Be sure the beveled or grooved
edge is facing upwards and that the post lines up with the
post of the lens, as shown in the following figure.
After the small fruit adaptor is installed, center the small
fruit specimen on the central ring. This small ring will hold
the fruit specimen in the correct position for
measurements to be taken. Now, place the reflector cone
over the specimen. You will notice that there are magnets
installed within the reflector cone to obtain a good fit.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
65
F-750 Instruction Manual 7/30/2015
Cleaning and Maintenance
Keep the lens of the F-750 clean and protected from dirt
and scratches. If the lens is dirty, unscrew the cover plate
and gently wipe the lens in a smooth straight motion with
rubbing alcohol and a clean nonabrasive cloth. Then, reaffix the cover plate. With normal use, the durable Gorilla
Glass lens should not scratch. If the lens requires rinsing
with water, ensure that the rubber bumper (that protects
water from leaking into the USB and SD card slot) is snugly
installed. The rubber bumper and the case of the F-750 can
be wiped clean with a cloth.
66
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Firmware Update Procedure
The firmware of the F-750 is updated by connecting the
instrument to a computer using a USB cable and running
the Firmware Update tool. The firmware update tool
includes details instructions for updating the firmware.
This firmware update tool is available online at
http://www.felixinstruments.com/Software/F750/FirmwareUpdateTool/index.htm
For questions concerning firmware updates, please
contact [email protected].
Tip: If you experience any issues with the firmware
updating process, “Bootloader Assist” mode can be
accessed by pressing and holding the ‘down’ arrow
while turning the device on. Follow the instructions
displayed on screen, and continue the firmware
updating process as usual.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
67
F-750 Instruction Manual 7/30/2015
Supporting Science
Fish Fat and Moisture Content
Downey, G. (1996). Non-invasive and non-destructive
percutaneous analysis of farmed salmon flesh by near
infra-red spectroscopy. Food Chem. 55:305–311.
 NIR interactance range 700-1100 nm
 6 sites on dorsal and ventral side of each fish
(measurement through skin and scales).
 Total of 294 spectra from all different sites.
 Reference chemical values of fat and moisture
were determined from excised flesh from the
different NIR measurement sites.
 Fat range: 2.3–23.0%
 Moisture range: 57.9–74.7%.
 Spectral measurements on the dorsal surface gave
lowest prediction errors (bias corrected) for fat
2.0% and moisture 1.45%.
Lee, M. H., Cavinato, A. G., Mayes, D. M., and Rasco, B.
A.(1992). Noninvasive short-wavelength near-infrared
spectroscopic method to estimate the crude lipid content
in the muscle of intact rainbow trout, J. Agric. Food Chem.
40:2178–2181.
 NIR interactance range 700-1050 nm
 52 frozen and thawed rainbow trout, Oncorhyncus
mykiss (measurement through skin).
 Weight range: 66.5–883 grams
 Local fat range: 2.0–13%
 Prediction error results for fat: 0.7 to 2.3%;
correlation coefficients ranged from 0.73 to 0.90
68
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
and varied with measurement locations on the fish
body.
Wold, J. P. and Isaksoon, T. (1997), Non-Destructive
Determination of Fat and Moisture in Whole Atlantic
Salmon by Near-Infrared Diffuse Spectroscopy. Journal of
Food Science, 62: 734–736.
 NIR interactance range 800-1100 nm (2 nm
increments/steps)
 Intact whole farmed Atlantic salmon
 Determine crude fat and moisture
 The method could be used for classification of
salmon into groups:
o Very low fat (less than 8%)
o Low fat (8–12%)
o Medium fat (12–16%)
o High fat (16–20%)
o Very high fat (greater than 20%)
Wine Grape Anthocyanin Content
Cozzolino, D., Parker, M., Dambergs, R. G., Herderich, M.
and Gishen, M. (2006), Chemometrics and visible-near
infrared spectroscopic monitoring of red wine
fermentation in a pilot scale. Biotechnol. Bioeng., 95:
1101–1107.
 NIR interactance range 400 -700 nm
 Began sampling on Day 0 of grape fermentation
 Samples taken after Day 0 showed a marked
increase in anthocyanin absorption around 540
nm
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
69
F-750 Instruction Manual 7/30/2015

Demonstrates extraction of these phenolic
pigments from grape skins into the wine during
fermentation.
Mango Model Optimal Wavelengths
Subedi, P., Walsh, K., Owens, G. (2007) Prediction of
mango eating quality at harvest using short-wave near
infrared spectrometry. Postharvest Biology and
Technology, 43: 326–334.
 Optimal wavelength range for DM modeling
provides the lowest RMSE.
o Start wavelength range of 700-880 nm
and end range of 920-1100 nm.
 Optimal wavelength range for TSS modeling:
o Start wavelength range of 740-850 nm
and end range of 950-1100 nm.
 Optimal wavelength range for Hunter b (skin
color) modeling:
o Start wavelength range of 760–870 nm
and end range of 1000-1100 nm.
 Optimal wavelength range for maturity score
modeling:
o Start wavelength range of 700-850 nm
and end range of 1000-1100 nm.
 The optimal wavelength region was similar for the
TSS, DM and Hunter b models, and different to
that
for
the
maturity
score
model.
70
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Kiwifruit
McGlone, V.A. (1998). Firmness, dry-matter and solublesolids assessment of postharvest kiwifruit by NIR
spectroscopy. Postharvest biology and technology (09255214), 13 (2), 131.
 NIR interactance range for Dry Matter and TSS
800-1100 nm.
 Known important carbohydrate bands at 880,
900–930 and 970 nm.
 Strong
water
band
at
958
nm.
Apple Dry Matter
McGlone, A., Jordan, R., Seelye, R., Clark, C. (2003). Drymatter – a better predictor of the post-storage soluble
solids in apples? Postharvest biology and technology, 28:
431–435.
 NIR interactance range for apple dry matter:
800-1000 nm.
Instrument Precision Optimization
Greensill, C., Walsh, K. (2000) Optimization of
Instrumentation Precision and Wavelength Resolution for
the Performance of NIR Calibrations of Sucrose in a WaterCellulose Matrix. Appl. Spectrosc. 54: 426-430.

NIR interactance range sucrose in water 7001050 nm
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
71
F-750 Instruction Manual 7/30/2015
Publications
Cozzolino, D., Parker, M., Dambergs, R. G., Herderich, M.
and Gishen, M. (2006), Chemometrics and visible-near
infrared spectroscopic monitoring of red wine
fermentation in a pilot scale. Biotechnol. Bioeng., 95:
1101–1107.
Downey, G. (1996). Non-invasive and non-destructive
percutanous analysis of farmed salmon flesh by near infrared spectroscopy. Food Chem. 55:305–311.
Greensill, C., Walsh, K. (2000) Optimization of
Instrumentation Precision and Wavelength Resolution for
the Performance of NIR Calibrations of Sucrose in a WaterCellulose Matrix. Appl. Spectrosc. 54: 426-430.
Lee, M. H., Cavinato, A. G., Mayes, D. M., and Rasco, B. A.
(1992). Noninvasive short-wavelength near-infrared
spectroscopic method to estimate the crude lipid content
in the muscle of intact rainbow trout, J. Agric. Food Chem.
40:2178–2181.
McGlone, V.A. (1998). Firmness, dry-matter and solublesolids assessment of postharvest kiwifruit by NIR
spectroscopy. Postharvest biology and technology (09255214), 13 (2), 131.
McGlone, A., Jordan, R., Seelye, R., Clark, C. (2003). Drymatter – a better predictor of the post-storage soluble
solids in apples? Postharvest biology and technology, 28:
431–435.
72
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Savitzky, A., Golay, M. (1964) Smoothing and
Differentiation of Data by Simplified Least Squares
Procedures. Analytical Chemistry, 36(8) 1627–1639.
Sjostrom, M., Wold, S. (1983) A multivariate calibration
problem in analytical chemistry solved by partial leastsquare models in latent variables. Analytica Chimica Acta,
(150) 61-70.
Subedi, P. (2007) Prediction of mango eating quality at
harvest using short-wave near infrared spectrometry.
Postharvest biology and technology, 43 (3), 326.
Subedi, P., Walsh, K. (2011) Assessment of sugar and
starch in intact banana and mango fruit by SWNIR
spectroscopy. Postharvest Biology and Technology
Subedi, P., Walsh, K. (2009) Technologies for assessing
fruit quantity and quality: maturity, pigmentation, dry
matter content, firmness. Mango Encylopaedia, 2(10): 139.
Subedi, P., Walsh, K., Hopkins, D. (2012) Assessment of
titratable acidity in fruit using short wave near infrared
spectroscopy. Part A: establishing a detection limit based
on model solutions. Near Infrared Spectrosc., (20) 449457.
Subedi, P., Walsh, K., Hopkins, D. (2012) Assessment of
titratable acidity in fruit using short wave near infrared
spectroscopy. Part B: intact fruit studies. Near Infrared
Spectrosc., (20) 459-463.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
73
F-750 Instruction Manual 7/30/2015
Subedi, P., Walsh, K., Owens, G. (2007) Prediction of
mango eating quality at harvest using short-wave near
infrared spectrometry. Postharvest Biology and
Technology, 43: 326–334.
Subedi, P., Walsh, K., Purdy, P. (2010) Determination of
Optimum Maturity Stages of Mangoes Using Fruit Spectral
Signatures, China Int Mango Conf 1-12.
Walsh, K., Guthrie, J., Burney, J. (2000) Aust. J Application
of commercially available, low-cost, miniaturized NIR
spectrometers to the assessment of the sugar content of
intact fruit. Plant Physiol, 27: 1175-1186.
Walsh, K., Long, R., Middelton, S. (2007) Use of near infrared spectroscopy in evaluation of source-sink
manipulation to increase the soluble sugar content of
stonefruit. Journal of Horticultural Science &
Biotechnology, (82:2) 316–322.
Wold, J., Isaksoon, T. (1997) Non-Destructive
Determination of Fat and Moisture in Whole Atlantic
Salmon by Near-Infrared Diffuse Spectroscopy. Journal of
Food Science, 62: 734–736.
74
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
F-750 Vocabulary Definitions
Aroma: Comes from volatiles given off by the fruit as they
are detected by olfactory receptors when fruit is chewed
and swallowed.
Dry Matter (DM): The ratio of the water content to the dry
weight of the fruit. Dry matter is a function of the solids
and water the fruit has accumulated while growing on the
plant, and is used as an indicator of both taste and texture.
Explained Variance: Share of total variance which is
accounted for by the model. For example, an explained
variance of 90% implies that the model accounts for 90%
of variance in the data.
Flavor: is a combination of taste and aroma
Horizontal Dilution of Precision (HDoP): The HDoP
represents an estimate of how much error is introduced to
GPS readings by the relative position of the satellites to the
user. HDoP is measured on a scale from 1 to 20, with one
being the best, and therefore most accurate, and 20 being
the worst, with around 300 meters of error in any
direction. An HDoP of 1 to 4 is recommended if GPS
readings are going to be taken in the field.
Interactance Spectroscopy: A mode of NIR spectroscopy
where the light is directed into the specimen, interacts
with the sample, and is remitted from another location on
the specimen which is not receiving light itself. This
method is ideal for providing information on the internal
composition of the sample (see figure in Theory of
Operation section page 6).
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
75
F-750 Instruction Manual 7/30/2015
Model Builder File: The type of file used by the Model
Builder software which includes the Training Set,
Reference Values and ranges of spectra to build the model
used by the device.
Model Linearity: ‘Goodness of fit’ for the selected
number of principle components.
NIPALS (Non-linear Iterative Partial Least Squares):
Iterative computer implementation of PLS. PLS is a method
used by the F-750 to generate a mathematical model of
regression coefficients. PLS1 only predicts a single trait
with a single regression vector. PLS2 can be used to predict
multiple traits with a single regression vector. The F-750
will do PLS1 multiple times to predict multiple traits.
Prefix: This is a label placed on each measurement along
with the lot and sample number that may help with
organizing data.
Prediction Error: Reference values versus predicted
values for each specimen. Cross validation uses ‘leave
one out’ grouping by specimen number to simulate an
independent data set.
Reference Value: The known concentrations or value of
constituent (i.e. trait, quality, or property of interest).
This could be the use of a refractometer for total soluble
sugar or a scale to determine dry matter content.
Regression Coefficients: Mathematical significance of
each pixel, selected in spectra selection, which is used to
make predictions. Regression coefficients are a series of
76
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
weighted values generated by PLS. When these are applied
to a scan of an unknown sample, a prediction of
concentration is made. These values are the core of the
mathematical model. PLS takes multiple scans and
combines it with Reference Values to generate a
regression vector specific to a single property (such as dry
matter). By taking the dot product of the regression vector
and a scan of an unknown sample, a concentration can be
determined. These are stored in a single row wise vector,
there is 1 element in the vector for every pixel used the
regression.
Root Mean Squared Error (RMSE): Average or percent
uncertainty (error) of any individual measurement. For
example, a dry matter model with a RMSE of 0.0123 has
an error of 1.23%DM or a 1.23% prediction error. An
acceptable amount of error is determined by the type of
property being predicted and the user application. It also
depends on the uncertainty of the reference method.
RMSE is treated as expected error with first standard
deviation or 68% of measurements.
Root Mean Squared Error Calibration (RMSEC):
Measurement of goodness of fit between data and
calibration model (training set). This is essentially the
calibration for training set error prediction.
Root Mean Squared Error Cross Validation (RMSECV): The
error of cross validation of training set prediction. One
specimen is taken out, the model rebuilt and that
specimen is predicted using the model.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
77
F-750 Instruction Manual 7/30/2015
Root Mean Squared Error Prediction (RMSEP): A measure
of true prediction, measured on real fruit and compared to
reference values of an independent data set.
Savitzky–Golay Coefficients: A mathematical tool used to
smooth spectra and to generate first or second
derivatives. This is useful as it removes noise and the
derivative removes baseline offset error. Refer to Savitzky
and Golay (1964) for more information.
Scores Plot: Displays how close each spectra is to the
principle components used to create the model. Distinct
groups of points indicate “like” specimen, such as when
viewing two different temperatures.
SSC: Soluble sugar content
Taste: The amount of sugars and acids in the fruit, as they
are detected by taste buds on the tongue.
Training Set: The set of all of scans that will be used for the
PLS regression.
X population: All of the spectra stored in a row rise matrix.
Y population: The Reference Values determined by a
reference method stored in a single column vector.
78
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Technical Support
If you have a question about the F-750, first look in the F750 Instruction Manual. There is also online support
available
for
the
F-750
at
http://felixinstruments.com/support. If you cannot find
the answer, you can contact a Technical Support
Representative located in your country. Felix Instruments
is committed to provide customers with high quality,
timely
technical
support.
Technical
support
representatives are to answer your technical questions by
phone or by e-mail at [email protected].
Felix Instruments’ contact information:
Felix Instruments
1554 NE 3rd Ave
Camas, WA 98607 USA
Phone: 800-767-0119 (U.S. and Canada)
360-833-8835
Fax: 360-833-1914
Website: http://www.felixinstruments.com
E-mail: [email protected]
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
79
F-750 Instruction Manual 7/30/2015
Customer Service
Customer Service Representatives answer questions about
specifications and pricing, and sell all of the Felix
Instruments products. Customers sometimes find that
they need Felix Instruments to upgrade, recalibrate or
repair their system. In order for Felix Instruments to offer
these services, the customer must first contact us and
obtain a Return Merchandise Authorization (RMA)
number. Please contact a customer service representative
for specific instructions when returning a product.
80
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Frequently Asked Questions
1. What can the F-50 Produce Quality Meter
measure?
The F-750 Produce Quality Meter is typically used
to measure the concentration of dry matter, total
soluble solids, chlorophyll, titratable acids, etc. in
fruits and other produce. However, the F-750 can
be used to measure any substance with a distinct
correlation to light in the 285-1200 nm range. This
includes everything from steel quality to the liquid
content of fish or dry matter in cheese.
2. How do I recharge the batteries of the F-750?
To recharge the batteries, peel off the rubber
bumper from the F-750. Then twist to remove the
battery compartment cap (located on the bottom
of the F-750). Next, remove the batteries from the
F-750 to place them in the battery charger.
Charged batteries should be inserted into the F750 positive (+) side first.
3. If the instrument locks up can I reset it without
removing the batteries?
Yes, press and hold the power button until you
hear a beep. This will cause the instrument to
reset. If you have held the power button for longer
than ten 10 seconds, and the instrument will not
restart, you may need to remove, then re-insert
the batteries.
4. When I power on my F-750, I receive a message
which reads “SD Card Corrupted,” What should I
do?
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
81
F-750 Instruction Manual 7/30/2015
This message indicates that an error has occurred
with the SD card. To check this, insert the SD card
into your computer. Go to your computer folder,
then right click on the SD card icon, and click on
properties. Click on the tools tab, and then the
option for check drive for errors. Click Check Now
and if you are given disk options, check both boxes
to fix errors, and to recover bad sectors. Your SD
card should now work if re-inserted into your F750.
5. I started a scan and it will not stop taking scans.
How do I stop it?
If you want to stop the device from taking
consecutive scans during the process, the
opportunity to stop the next scan comes after the
measurement data screen displays. When the
screen displays after the measurement data
screen, press any arrow key. If you have the Delay
Between Measurements set to 0, you can hold
down any arrow key once the measurement data
screen displays until it disappears. After the
measurement data screen displays and an arrow
key is pressed, you will get a display that gives you
the option to press the left arrow to abort or the
right arrow to continue taking continuous scans.
6. I am having trouble updating the firmware on my
device, what can I do?
If you experience any issues with the firmware
updating process, “Bootloader Assist” mode can
be accessed by pressing and holding the ‘down’
arrow while turning the device on. Follow the
instructions displayed on screen, and continue the
firmware updating process as usual.
82
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
7. How many specimen should I use to create a
model?
As a general rule, we recommend a minimum of
200 specimen to build a model. The upper limit for
number of specimen in a model can be detected
by monitoring the RMSECV. When there is no
longer a change in RMSECV by addition of more
samples, the upper limit is reached. As the number
of samples increased, the RMSECV should
decrease until you see diminishing returns. At this
point you have sufficiently sampled the variance in
the population. For a proof of concept model, 2030 specimen should be sufficient.
8. How do I change the name of a model after I have
already saved it to my F750?
The only way you can set the name of the model is
when you Save As… to the SD card from Model
Builder. If you would like to change the name of a
model, you would have to load the model into
Model Builder and Save As… with the desired
name.
9. Why is the scale of the Savitzky–Golay second
derivative off by a factor of 2 in intensity?
The F750 and Model Builder software were
designed to be compatible with the original
NIRvana. As such, the coefficients are identical to
unscrambler V8 and below.
10. I am having a problem. I have read the manual and
the FAQ’s and I still don’t know what to do. What
do I do now?
If you have a question about the F-750, first look
in the F-750 Instruction Manual. There is also
online support available for the F-750 at
1554 NE 3rd Ave, Camas, WA 98607, USA
83
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
http://felixinstruments.com/support.
If you
cannot find the answer, you can contact a
Technical Support Representative located in your
country. Felix Instruments is committed to provide
customers with high quality, timely technical
support. Technical support representatives are to
answer your technical questions by phone or by email at [email protected].
84
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Felix Instruments Hardware Warranty
Important: Please Read!
Seller’s Warranty and Liability: Seller warrants new
equipment of its own manufacturing against defective
workmanship and materials for a period of one year, of a
single shift operation, from date of receipt of equipment the results of ordinary wear and tear, neglect, misuse,
accident and excessive deterioration due to corrosion
from any cause is not to be considered a defect. Any
defect must be called to the attention of Felix Instruments,
Camas, Washington, USA, in writing, within 90 days after
receipt of the unit.
Seller’s liability for defective parts is limited to the repair
or replacement of any part of the instrument without
charge, if Felix Instruments’ examination discloses that
part to have been defective in material or workmanship,
and in no event shall exceed the furnishing of replacement
parts F.O.B. the factory where originally manufactured. No
equipment may be repaired or altered by anyone not
authorized by Felix Instruments.
Material and equipment covered hereby, which is not
manufactured by Seller, is to be covered only by the
warranty of its manufacturer. Seller shall not be liable to
the Buyer for loss, damage, or injury to persons (including
death), or to property or things, whatsoever, including, but
without limitation, products processed by the use of the
equipment; or for damages of any kind or nature
(including, but without limitation, loss of anticipated
profits), occasioned by or arising out of installation,
operation, use, misuse, nonuse, repair, or replacement of
said material and equipment, or out of the use of any
1554 NE 3rd Ave, Camas, WA 98607, USA
85
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
method or process for which the same may be employed.
The purchaser is to pack, ship, or deliver the instrument to
Felix Instruments, in Camas, Washington, USA, within 30
days after Felix Instruments has received written notice of
the defect at the customer’s expense. No other
arrangements may be made unless otherwise approved in
writing by Felix Instruments.
The use of this equipment constitutes Buyer’s acceptance
of the terms set forth in this warranty. There are no
understandings, representations, or warranties of any
kind, express, implied, statutory, or otherwise (including,
but without limitation, the implied warranties of
merchantability and fitness for a particular purpose), not
expressly set forth herein.
86
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
Appendix I: Mango Model Building
Standard Operating Procedure (SOP)
The following procedure outlines the process of creating
a model on the F-750 device for dry matter and TSS (brix)
with mangos.
I. Getting prepared for model building
Equipment needed:
 Environmental chamber at temperature
 Fruit
 Permanent Marker(s)
1. Select specimen of fruit with a wide range of
maturity to be used in training set creation and
label with a specimen number. (Note: if both
sides of the fruit are to be scanned label each
side as their own sample #, i.e.; mango 1 will be
samples 1 and 2, mango 2 will be samples 3 and 4
and so on.)
2. Determine the three temperatures to be used for
the training set (Minimum, Mid, and Maximum.)
It is best to choose a range of temperatures
encompassing the range which the
measurements will be taken in the field, for
example, 10°C, 20°C, 30°C.
3. Place the fruit in a temperature controlled
environmental chamber (a temperature
controlled room will work.)
4. After the fruit has been given enough time to
come to the pre-determined temperature
(generally about one hour.) It is now time to
begin taking measurements for the training set.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
87
F-750 Instruction Manual 7/30/2015
II. Preparing the F-750
Equipment needed:
 F-750
1. Turn on the F-750 and navigate to the Setup
Menu > Training Sets > Create New Training Set.
2. Enter a name to use for the new training set
using the up and down arrows to select
characters and press the right arrow to enter
(pressing the right arrow twice moves to the next
menu.)
3. Enter the number of specimens that will be used
to build the model (minimum of 10 specimens.)
Press the right arrow again to create the training
set file. Be aware this step can take a few
minutes, and very large training sets (5000+) can
take up to an hour to process.
4. After the training set has been created, press
right to enter the Browse Specimen Data menu.
Navigate to the first temperature: Minimum.
Now we will begin scanning each specimen. Be
careful not to remove too many specimens at a time
from the temperature controlled chamber as you
want the set temperature to remain constant across
the specimens.
88
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
III. Scanning Fruit
Equipment needed:
 Permanent Marker(s)
 F-750
1. Place the first specimen on the fruit adaptor of
the F-750 and use a permanent marker to mark
the area to be scanned.
2. On the F-750 with Specimen 1 highlighted on the
list, press the right arrow to collect the training
spectra. The instrument will indicate
“Processing”. Do not press the scan button to
gather training set spectra.
3. After each specimen has been scanned, you will
see a date and time included in the specimen file
name. (Note: Specimens can be re-recorded if a
mistake is made or the wrong
specimen/temperature is selected.)
4. Repeat for all specimens at the Minimum
Temperature.
5. After all specimens have been recorded at the
Minimum Temperature, place the specimens
back in the temperature controlled area to bring
the fruit to your next selected temperature: Mid
temperature.
6. After specimens have reached the desire Mid
temperature, the next set of temperature data is
ready to be collected.
7. If you have turned off the F-750 while waiting for
the specimens to reach their next temperature,
navigate to the Setup Menu > Training Sets >
Modify Existing Training Set and select the
training set you created previously.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
89
F-750 Instruction Manual 7/30/2015
8. Within your training set file go to Browse
Specimen Data and navigate to the Mid
Temperature folder. Place the first specimen on
the fruit adaptor and align with pervious san
location markings, and with Specimen 1
highlighted on the list press the right arrow to
collect the training spectra.
9. Repeat for all specimens at Mid temperature.
(Note: You will be using the same specimens for
all three temperature sets so the specimen must
be scanned in the same order and at the same
location as was done for the Minimum
Temperature.)
10. Repeat the previous steps for the Maximum
Temperature when the specimens are at the
selected temperature.
You are now finished creating your training set. The same
specimens used for training set creation will now be used
to collect reference values for building a predictive
model.
90
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
IV. Collecting Dry Matter Reference Data
Equipment Needed:





Paring knife
Cutting board
Circular corer
Microbalance
Food dehydrator(s)
1. Remove the fruit skin from the area which was
marked and scanned with the F-750 with a sharp
knife being careful to both fully remove the skin,
but also to remove as little meat as possible.
2. Push the 26mm circular corer to the seed of the
mango in the prepared area.
3. Remove the corer and slice the mango close to
the seed removing the core of sampled mango.
4. Trim the mango core on the seed side, so the
total height of the cylinder is 2cm.
5. Weigh the sample in a microbalance as soon as is
possible and record the weight to the nearest
.001 grams. The expected wet mass should be
around 10-12 g.
6. Dry the samples in a dehydrator for 48 hours set
at 63-65°C.
7. After 40 hours, re-weigh a few of the samples
and record the values. Replace the samples back
into the dehydrator.
8. Re-check the same samples a couple of hours
later and see if they have changed in weight, if
they have not you can assume they are fully
dehydrated.
9. Measure all samples one at a time in a
microbalance and record the values.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
91
F-750 Instruction Manual 7/30/2015
10. Divide the final (dry) weight from the initial (wet)
weight for each sample to determine the Dry
Matter.
IV. Collecting Brix/SSC Reference Data
Equipment Needed:





Paring knife
Cutting board
Circular corer
Digital Refractometer
Garlic Press/cheese cloth
1. Ensure fruit and refractometer have equilibrate
to the same temperature.
2. Remove the fruit skin from the area which was
marked and scanned with the F-750 with a sharp
knife being careful to both fully remove the skin,
but also to remove as little meat as possible.
3. Push the 26mm circular corer to the seed of the
mango in the prepared area.
4. Remove the corer and slice the mango close to
the seed removing the core of sampled mango.
5. Trim the mango core on the seed side, so the
total height of the cylinder is 2cm.
6. Load the sample core into the garlic press or
cheese cloth
7. Holding the press over the refractometer, apply
pressure until several drops of juices have
covered the quartz optical sensor of the
refractometer.
8. Press the scan button and record the value
92
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
9. Press the scan button again to ensure the
refractometer has given a stable reading
10. Wipe and dry the refractometer
11. Repeat process with the next specimen.
12. It is useful to occasionally check the zero of the
refractometer with DI water.
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
93
F-750 Instruction Manual 7/30/2015
F-750 Production Test Check Sheet
Date: ___________________
Technician Name: ___________________
Serial #: ___________________
Spectrometer #: ___________________
Firmware Version: ___________________
Spectrometer Pixel Coefficients
C0:
C1:
C2:
C3:
Spectrometer ADC Gain:
Reference Voltage:
94
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
F-750 Instruction Manual 7/30/2015
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com
95
F-750 Instruction Manual 7/30/2015
Warranty Registration Card
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835 Fax: (360) 833-1914 e-mail: [email protected] Web:
www.felixinstruments.com
PRODUCT REGISTRATION CARD
Please complete and return this form to Felix Instruments within 30 days to validate your Warranty on
Parts & Labor.
Registration
Information:
Your Name:____________________________________ Title:__________________
Company/University:___________________________________________________
Address:_____________________________________________________________
City:____________________________ State:__________ Zip:__________________
Country:__________________________Email_______________________________
Phone:______________________________ Fax:____________________________
Felix Instruments Serial Number(s):__________________________________________________
Purchase Date:_____________________Purchase Price:_____________________
-----------------------------------------------------------------------------------------------------------------FOLD ON DOTTED LINE
Your opinions will help improve our service. Please answer the following questions.
1. What was the basis of your product selection?
□ Representative Recommendation
□ Product Features
□ Technical Specifications
□ Warranty
□ Other _________________________________________
□ Price
□ Product Design
□ Brand Name
□ Service
2. What other competing brands did you consider? __________________________
3. Where did you first learn of this product?
□ Advertisement in ________________________________
□ Friend/Colleague
□ Other _________________________________________
□ Representative
□ Exhibit
4. Who selected this product?
□ I did
□ University Department
□ Other _________________________________________
□ Research Group
□ Purchasing
5. Comments/Suggestions:
______________________________________________________________________
______________________________________________________________________
96
1554 NE 3rd Ave, Camas, WA 98607, USA
Phone: (360) 833-8835
[email protected]
www.felixinstruments.com